Re: [boost] [Serliazation] Thread-safety again

From: Robert Ramey (ramey_at_[hidden])
Date: 2006-12-17 17:20:20

Emil Dotchevski wrote:

>>>> As far as "deadstripping" code - this has indeed been a big problem
>>>> - especially in release mode. I've referred to it as overzealous
>>>> optimization.
>>
>>> It isn't overzealous -- it's within the specifications of the
>>> standard, whether you like it or not.
>>
>> Chapter and verse? (Given the above argument that dynamic
>> initialization must happen first.)
>
> The standard does not deal with deadstripping but to me it's obvious
> that if A) the compiler is not required to initialize a
> namespace-scope object until a function from its translation unit is
> called, and B) no function from that translation unit is called, the
> obvious conclusion is that the object will never be initialized and
> therefore can be safely deadstripped.

If I recal, the "dead stripping" occured even if a function was called from 
the class.

test_exported calls the constructor for a derived class (via new)  but 
doesn't explicitly call any "serialize" function.  WIthout extra "help", the 
program won't link in release mode.  I called this  "overzealous 
optimization". I'm not sure we're referring to the same thing here.  I 
solved it by marking the dropped functions with "_export" so that the linker 
wouldn't know they were never called.  Again, I'm not sure we're referring 
to the same thing here.  Then there's the question of an inline constructor. 
Its not at all clear to me what translation unit - if any it belongs.  Note 
that the problem I've been referring to even occurs in the absense of export 
so - I don't think this (at least this particular problem) is a C++ 
standards issue.  Except perhaps to say the C++ doesn't even address things 
like dynamic export of functions which is very common in windows programs at 
least.

Robert Ramey